Agent Workflow Designer: AI-Driven Workflow Auditing and HTML Report Generator
中文What
A structured system prompt and workflow that acts as an advisory AI agent. Through a 10-15 minute single-question interview, it audits a user's daily job tasks, identifies high-value AI automation opportunities (classified from L1 to L5 automation levels), and generates a comprehensive, visually styled HTML diagnostic report containing a 7-day startup plan and a 30-day roadmap.
How
Copy the prompt from SKILL.md into any platform supporting custom system prompts (e.g., Claude, ChatGPT, Dify, Coze). Run the session to undergo an 8-step interview: usage confirmation, profile gathering, workflow mapping, pain-point identification, automation analysis, workflow design (including inputs, steps, outputs, and human-in-the-loop audit points), prioritization, and final HTML report rendering.
Why
Generic advice like 'use AI to save time' is rarely actionable. This tool maps real-world work before applying technology, adheres to a 'semi-automatic before fully-automatic' philosophy, explicitly warns against automating low-value tasks, and packages the final analysis into a structured HTML report rather than a transient chat transcript.
Other
Categorizes automation into 5 levels: L1 (assisted draft generation), L2 (multi-step semi-automated with human audits), L3 (multi-tool execution), L4 (event-driven triggers), and L5 (autonomous execution in safe sandboxes). It is MIT licensed and compatible with major LLMs including Claude, GPT, and custom Agent orchestrators.